bl1_axpysmt.c File Reference

(r)

Functions
void	bl1_saxpysmt (trans1_t trans, int m, int n, float *alpha0, float *alpha1, float *a, int a_rs, int a_cs, float *beta, float *b, int b_rs, int b_cs)
void	bl1_daxpysmt (trans1_t trans, int m, int n, double *alpha0, double *alpha1, double *a, int a_rs, int a_cs, double *beta, double *b, int b_rs, int b_cs)
void	bl1_caxpysmt (trans1_t trans, int m, int n, scomplex *alpha0, scomplex *alpha1, scomplex *a, int a_rs, int a_cs, scomplex *beta, scomplex *b, int b_rs, int b_cs)
void	bl1_zaxpysmt (trans1_t trans, int m, int n, dcomplex *alpha0, dcomplex *alpha1, dcomplex *a, int a_rs, int a_cs, dcomplex *beta, dcomplex *b, int b_rs, int b_cs)

Function Documentation

bl1_caxpysmt()

void bl1_caxpysmt	(	trans1_t	trans,
		int	m,
		int	n,
		scomplex *	alpha0,
		scomplex *	alpha1,
		scomplex *	a,
		int	a_rs,
		int	a_cs,
		scomplex *	beta,
		scomplex *	b,
		int	b_rs,
		int	b_cs
	)

{
    scomplex* a_begin;
    scomplex* b_begin;
    scomplex* a_temp;
    scomplex  alpha_prod;
    int       inca_temp;
    int       lda, inca;
    int       ldb, incb;
    int       n_iter;
    int       n_elem;
    int       j;
 
    // Return early if possible.
    if ( bl1_zero_dim2( m, n ) ) return;
 
    alpha_prod.real = alpha0->real * alpha1->real - alpha0->imag * alpha1->imag;
    alpha_prod.imag = alpha0->real * alpha1->imag + alpha0->imag * alpha1->real;
 
    // Handle cases where A and B are vectors to ensure that the underlying axpy
    // gets invoked only once.
    if ( bl1_is_vector( m, n ) )
    {
        // Initialize with values appropriate for vectors.
        n_iter = 1;
        n_elem = bl1_vector_dim( m, n );
        lda    = 1; // multiplied by zero when n_iter == 1; not needed.
        inca   = bl1_vector_inc( trans,             m, n, a_rs, a_cs );
        ldb    = 1; // multiplied by zero when n_iter == 1; not needed.
        incb   = bl1_vector_inc( BLIS1_NO_TRANSPOSE, m, n, b_rs, b_cs );
    }
    else // matrix case
    {
        // Initialize with optimal values for column-major storage.
        n_iter = n;
        n_elem = m;
        lda    = a_cs;
        inca   = a_rs;
        ldb    = b_cs;
        incb   = b_rs;
        
        // Handle the transposition of A.
        if ( bl1_does_trans( trans ) )
        {
            bl1_swap_ints( lda, inca );
        }
 
        // An optimization: if B is row-major and if A is effectively row-major
        // after a possible transposition, then let's access the matrices by rows
        // instead of by columns for increased spatial locality.
        if ( bl1_is_row_storage( b_rs, b_cs ) )
        {
            if ( ( bl1_is_col_storage( a_rs, a_cs ) && bl1_does_trans( trans ) ) ||
                 ( bl1_is_row_storage( a_rs, a_cs ) && bl1_does_notrans( trans ) ) )
            {
                bl1_swap_ints( n_iter, n_elem );
                bl1_swap_ints( lda, inca );
                bl1_swap_ints( ldb, incb );
            }
        }
    }
 
    if ( bl1_does_conj( trans ) )
    {
        conj1_t conj = bl1_proj_trans1_to_conj( trans );
 
        a_temp = bl1_callocv( n_elem );
        inca_temp = 1;
 
        for ( j = 0; j < n_iter; j++ )
        {
            a_begin = a + j*lda;
            b_begin = b + j*ldb;
 
            bl1_ccopyv( conj,
                        n_elem,
                        a_begin, inca,
                        a_temp,  inca_temp );
 
            bl1_cscal( n_elem,
                       beta,
                       b_begin, incb );
 
            bl1_caxpy( n_elem,
                       &alpha_prod,
                       a_temp,  inca_temp, 
                       b_begin, incb );
        }
    
        bl1_cfree( a_temp );
    }
    else // if ( !bl1_does_conj( trans ) )
    {
        for ( j = 0; j < n_iter; j++ )
        {
            a_begin = a + j*lda;
            b_begin = b + j*ldb;
 
            bl1_cscal( n_elem,
                       beta,
                       b_begin, incb );
 
            bl1_caxpy( n_elem,
                       &alpha_prod,
                       a_begin, inca, 
                       b_begin, incb );
        }
    }
}

References alpha1, bl1_callocv(), bl1_caxpy(), bl1_ccopyv(), bl1_cfree(), bl1_cscal(), bl1_does_conj(), bl1_does_notrans(), bl1_does_trans(), bl1_is_col_storage(), bl1_is_row_storage(), bl1_is_vector(), bl1_proj_trans1_to_conj(), bl1_vector_dim(), bl1_vector_inc(), bl1_zero_dim2(), BLIS1_NO_TRANSPOSE, scomplex::imag, and scomplex::real.

Referenced by FLA_Axpys_external().

bl1_daxpysmt()

void bl1_daxpysmt	(	trans1_t	trans,
		int	m,
		int	n,
		double *	alpha0,
		double *	alpha1,
		double *	a,
		int	a_rs,
		int	a_cs,
		double *	beta,
		double *	b,
		int	b_rs,
		int	b_cs
	)

{
    double*   a_begin;
    double*   b_begin;
    double    alpha_prod;
    int       lda, inca;
    int       ldb, incb;
    int       n_iter;
    int       n_elem;
    int       j;
 
    // Return early if possible.
    if ( bl1_zero_dim2( m, n ) ) return;
 
    alpha_prod = (*alpha0) * (*alpha1);
 
    // Handle cases where A and B are vectors to ensure that the underlying axpy
    // gets invoked only once.
    if ( bl1_is_vector( m, n ) )
    {
        // Initialize with values appropriate for vectors.
        n_iter = 1;
        n_elem = bl1_vector_dim( m, n );
        lda    = 1; // multiplied by zero when n_iter == 1; not needed.
        inca   = bl1_vector_inc( trans,             m, n, a_rs, a_cs );
        ldb    = 1; // multiplied by zero when n_iter == 1; not needed.
        incb   = bl1_vector_inc( BLIS1_NO_TRANSPOSE, m, n, b_rs, b_cs );
    }
    else // matrix case
    {
        // Initialize with optimal values for column-major storage.
        n_iter = n;
        n_elem = m;
        lda    = a_cs;
        inca   = a_rs;
        ldb    = b_cs;
        incb   = b_rs;
        
        // Handle the transposition of A.
        if ( bl1_does_trans( trans ) )
        {
            bl1_swap_ints( lda, inca );
        }
 
        // An optimization: if B is row-major and if A is effectively row-major
        // after a possible transposition, then let's access the matrices by rows
        // instead of by columns for increased spatial locality.
        if ( bl1_is_row_storage( b_rs, b_cs ) )
        {
            if ( ( bl1_is_col_storage( a_rs, a_cs ) && bl1_does_trans( trans ) ) ||
                 ( bl1_is_row_storage( a_rs, a_cs ) && bl1_does_notrans( trans ) ) )
            {
                bl1_swap_ints( n_iter, n_elem );
                bl1_swap_ints( lda, inca );
                bl1_swap_ints( ldb, incb );
            }
        }
    }
 
    for ( j = 0; j < n_iter; j++ )
    {
        a_begin = a + j*lda;
        b_begin = b + j*ldb;
 
        bl1_dscal( n_elem,
                   beta,
                   b_begin, incb );
 
        bl1_daxpy( n_elem,
                   &alpha_prod,
                   a_begin, inca, 
                   b_begin, incb );
    }
}

References bl1_daxpy(), bl1_does_notrans(), bl1_does_trans(), bl1_dscal(), bl1_is_col_storage(), bl1_is_row_storage(), bl1_is_vector(), bl1_vector_dim(), bl1_vector_inc(), bl1_zero_dim2(), and BLIS1_NO_TRANSPOSE.

Referenced by FLA_Axpys_external().

bl1_saxpysmt()

void bl1_saxpysmt	(	trans1_t	trans,
		int	m,
		int	n,
		float *	alpha0,
		float *	alpha1,
		float *	a,
		int	a_rs,
		int	a_cs,
		float *	beta,
		float *	b,
		int	b_rs,
		int	b_cs
	)

{
    float*    a_begin;
    float*    b_begin;
    float     alpha_prod;
    int       lda, inca;
    int       ldb, incb;
    int       n_iter;
    int       n_elem;
    int       j;
 
    // Return early if possible.
    if ( bl1_zero_dim2( m, n ) ) return;
 
    alpha_prod = (*alpha0) * (*alpha1);
 
    // Handle cases where A and B are vectors to ensure that the underlying axpy
    // gets invoked only once.
    if ( bl1_is_vector( m, n ) )
    {
        // Initialize with values appropriate for vectors.
        n_iter = 1;
        n_elem = bl1_vector_dim( m, n );
        lda    = 1; // multiplied by zero when n_iter == 1; not needed.
        inca   = bl1_vector_inc( trans,             m, n, a_rs, a_cs );
        ldb    = 1; // multiplied by zero when n_iter == 1; not needed.
        incb   = bl1_vector_inc( BLIS1_NO_TRANSPOSE, m, n, b_rs, b_cs );
    }
    else // matrix case
    {
        // Initialize with optimal values for column-major storage.
        n_iter = n;
        n_elem = m;
        lda    = a_cs;
        inca   = a_rs;
        ldb    = b_cs;
        incb   = b_rs;
        
        // Handle the transposition of A.
        if ( bl1_does_trans( trans ) )
        {
            bl1_swap_ints( lda, inca );
        }
 
        // An optimization: if B is row-major and if A is effectively row-major
        // after a possible transposition, then let's access the matrices by rows
        // instead of by columns for increased spatial locality.
        if ( bl1_is_row_storage( b_rs, b_cs ) )
        {
            if ( ( bl1_is_col_storage( a_rs, a_cs ) && bl1_does_trans( trans ) ) ||
                 ( bl1_is_row_storage( a_rs, a_cs ) && bl1_does_notrans( trans ) ) )
            {
                bl1_swap_ints( n_iter, n_elem );
                bl1_swap_ints( lda, inca );
                bl1_swap_ints( ldb, incb );
            }
        }
    }
 
    for ( j = 0; j < n_iter; j++ )
    {
        a_begin = a + j*lda;
        b_begin = b + j*ldb;
 
        bl1_sscal( n_elem,
                   beta,
                   b_begin, incb );
 
        bl1_saxpy( n_elem,
                   &alpha_prod,
                   a_begin, inca, 
                   b_begin, incb );
    }
}

References bl1_does_notrans(), bl1_does_trans(), bl1_is_col_storage(), bl1_is_row_storage(), bl1_is_vector(), bl1_saxpy(), bl1_sscal(), bl1_vector_dim(), bl1_vector_inc(), bl1_zero_dim2(), and BLIS1_NO_TRANSPOSE.

Referenced by FLA_Axpys_external().

bl1_zaxpysmt()

void bl1_zaxpysmt	(	trans1_t	trans,
		int	m,
		int	n,
		dcomplex *	alpha0,
		dcomplex *	alpha1,
		dcomplex *	a,
		int	a_rs,
		int	a_cs,
		dcomplex *	beta,
		dcomplex *	b,
		int	b_rs,
		int	b_cs
	)

{
    dcomplex* a_begin;
    dcomplex* b_begin;
    dcomplex* a_temp;
    dcomplex  alpha_prod;
    int       inca_temp;
    int       lda, inca;
    int       ldb, incb;
    int       n_iter;
    int       n_elem;
    int       j;
 
    // Return early if possible.
    if ( bl1_zero_dim2( m, n ) ) return;
 
    alpha_prod.real = alpha0->real * alpha1->real - alpha0->imag * alpha1->imag;
    alpha_prod.imag = alpha0->real * alpha1->imag + alpha0->imag * alpha1->real;
 
    // Handle cases where A and B are vectors to ensure that the underlying axpy
    // gets invoked only once.
    if ( bl1_is_vector( m, n ) )
    {
        // Initialize with values appropriate for vectors.
        n_iter = 1;
        n_elem = bl1_vector_dim( m, n );
        lda    = 1; // multiplied by zero when n_iter == 1; not needed.
        inca   = bl1_vector_inc( trans,             m, n, a_rs, a_cs );
        ldb    = 1; // multiplied by zero when n_iter == 1; not needed.
        incb   = bl1_vector_inc( BLIS1_NO_TRANSPOSE, m, n, b_rs, b_cs );
    }
    else // matrix case
    {
        // Initialize with optimal values for column-major storage.
        n_iter = n;
        n_elem = m;
        lda    = a_cs;
        inca   = a_rs;
        ldb    = b_cs;
        incb   = b_rs;
        
        // Handle the transposition of A.
        if ( bl1_does_trans( trans ) )
        {
            bl1_swap_ints( lda, inca );
        }
 
        // An optimization: if B is row-major and if A is effectively row-major
        // after a possible transposition, then let's access the matrices by rows
        // instead of by columns for increased spatial locality.
        if ( bl1_is_row_storage( b_rs, b_cs ) )
        {
            if ( ( bl1_is_col_storage( a_rs, a_cs ) && bl1_does_trans( trans ) ) ||
                 ( bl1_is_row_storage( a_rs, a_cs ) && bl1_does_notrans( trans ) ) )
            {
                bl1_swap_ints( n_iter, n_elem );
                bl1_swap_ints( lda, inca );
                bl1_swap_ints( ldb, incb );
            }
        }
    }
 
    if ( bl1_does_conj( trans ) )
    {
        conj1_t conj = bl1_proj_trans1_to_conj( trans );
 
        a_temp = bl1_zallocv( n_elem );
        inca_temp = 1;
 
        for ( j = 0; j < n_iter; j++ )
        {
            a_begin = a + j*lda;
            b_begin = b + j*ldb;
 
            bl1_zcopyv( conj,
                        n_elem,
                        a_begin, inca,
                        a_temp,  inca_temp );
 
            bl1_zscal( n_elem,
                       beta,
                       b_begin, incb );
 
            bl1_zaxpy( n_elem,
                       &alpha_prod,
                       a_temp,  inca_temp, 
                       b_begin, incb );
        }
    
        bl1_zfree( a_temp );
    }
    else // if ( !bl1_does_conj( trans ) )
    {
        for ( j = 0; j < n_iter; j++ )
        {
            a_begin = a + j*lda;
            b_begin = b + j*ldb;
 
            bl1_zscal( n_elem,
                       beta,
                       b_begin, incb );
 
            bl1_zaxpy( n_elem,
                       &alpha_prod,
                       a_begin, inca, 
                       b_begin, incb );
        }
    }
}

References alpha1, bl1_does_conj(), bl1_does_notrans(), bl1_does_trans(), bl1_is_col_storage(), bl1_is_row_storage(), bl1_is_vector(), bl1_proj_trans1_to_conj(), bl1_vector_dim(), bl1_vector_inc(), bl1_zallocv(), bl1_zaxpy(), bl1_zcopyv(), bl1_zero_dim2(), bl1_zfree(), bl1_zscal(), BLIS1_NO_TRANSPOSE, dcomplex::imag, and dcomplex::real.

Referenced by FLA_Axpys_external().